Claude Mods: Flipping the Harness Inside Out
Latent Space Dan McAteer ● Covered by 2 sources
Claude Mods lets users customize Claude Code by installing JavaScript or TypeScript “mods” that run inside Claude Code to change how the agent interacts with the human operator during prompting, tool calls, and rendering. One built-in mod called You Should Know was turned on to catch 45 files with uncommitted edits after an apparently finished UI-editing task. As a result, the agent harness can be “reversed” into a programmable Attention Interface where the human and agent jointly shape what gets shown and what decisions require human input, alongside tradeoffs like noise, higher token cost, and permissions/trust concerns.
Why it matters
Claude Mods is an early form of the Attention Interface. It flips the agent harness paradigm from the harness facing the machine, to the harness facing the human.